Credit rating with a monotonicity-constrained support vector machine model

Deciding whether borrowers can fulfill their obligations is a major issue for financial institutions, and while various credit rating models have been developed to help achieve this, they cannot reflect the domain knowledge of human experts. This paper proposes a new rating model based on a support vector machine with monotonicity constraints derived from the prior knowledge of financial experts. Experiments conducted on real-world data sets show that the proposed method, not only data driven but also domain knowledge oriented, can help correct the loss of monotonicity in data occurring during the collecting process, and performs better than the conventional counterpart.     

A level set method based on the Bayesian risk for medical image segmentation

This paper proposes an alternative criterion derived from the Bayesian risk classification error for image segmentation. The proposed model introduces a region-based force determined through the difference of the posterior image densities for the different classes, a term based on the prior probability derived from Kullback-Leibler information number, and a regularity term adopted to avoid the generation of excessively irregular and small segmented regions. Compared with other level set methods, the proposed approach relies on the optimum decision of pixel classification and the estimates of prior probabilities; thus the approach has more reliability in theory and practice. Experiments show that the proposed approach is able to extract the complicated shapes of targets and robust for various types of medical images. Moreover, the algorithm can be easily extendable for multiphase segmentation.     

中间件的优先队列技术研究

在三层体系结构（客户端－中间件－服务器端）中,中间件作为客户端和服务端的中间层,起到至关重要的作用,优先队列是指在中间件的应用程序需要处理多个任务时,按任务的轻重顺序执行,使用户的平均等待时间最短,本文讨论了用二叉堆算法实现优先队列。     

A new learning method using prior information of neural networks

In this paper, we present a new learning method using prior information for three-layer neural networks. Usually when neural networks are used for identification of systems, all of their weights are trained independently, without considering interrelated weights values. Thus, the training results are usually not good. The reason for this in that each parameter has its influence on others during learning. To overcome this problem, we first give an exact mathematical equation that describes the relation between weight values given a set of data conveying prior information. The we present a new learning method that trains part of the weights and calculates the others using these exact mathematical equations. This method often a priori keeps the given mathematical structure exactly the same during learning; in other words, training is done so that the network follows a predetermined trajectory. Numerical computer simulation results are provided to support this approach. This work was presented, in part, at the Fourth International Symposium on Artificial Life and Robotics, Oita, Japan, January 19–22, 1999.     

基于先验形状信息的水平集图像分割

针对现有水平集方法对于具有强噪声或弱边界的目标进行分割时存在的问题,提出了一种基于形状先验的图像分割方法.该模型采用变分水平集方法,融合了区域特征和边界轮廓特征,并通过相似性匹配选择最佳先验形状.该模型不仅对具有强噪声和弱边界的复杂图像具有较好的分割效果,而且有效地解决了曲线演化的初始轮廓的确定问题.与传统方法进行对比实验,结果表明,该方法具有较好的分割效果和较高的准确率.     

煮糖过程的混合建模

煮糖过程是一个具有严重的非线性和不稳定性过程，过程的主要非线性表现在晶体生长率。本论文采用两种方法来描述晶体生长率模型的设计，第一种方法是通过采用非线性规划最优化方法(NLP)决定晶体生长率经验公式的参数；第二种方法是一种新颖的建模策略，它把晶体生长率作为非线性逼近器的人工神经网络(ANN)与由蔗糖晶体质量平衡所表述的先前既定的知识相结合。最初结果显示第一种类型的模型能够较好地执行局部匹配，而第二种具有更大的灵活性。     

Tools and Methods for the Improvement of Quality

In developing quality-control procedures, a step-loss function has been used implicitly or explicitly to describe consumer perceptions about product quality. A quadratic loss function has been suggested by Taguchi as an alternative to the step-loss function in measuring the loss due to imperfect product quality (cost of acceptance). In this article, Bayesian analyses of the known-standard-deviation acceptance-sampling problem are described for both the step and quadratic loss functions with three cost components—cost of inspection, cost of acceptance, and cost of rejection. A normal prior distribution is used for the lot mean. Efficient procedures for finding minimum expected cost procedures are given. For a particular example, comparisons are made of how optimal sampling plans and costs computed under the two cost structures change as the form of the prior distribution and misspecification of its mean and variance are varied. Sensitivity analyses for both cost functions show that the optimal sampling plan is robust with respect to the form of the prior distribution, as well as to misspecification of its mean and variance, if the tail specification reasonably approximates that of a normal distribution.     

Bayesian assessment of the characteristic concrete compressive strength using combined vague-informative priors

Bayesian statistics can be used in order to determine the characteristic value of an unknown distribution based on a limited number of test samples. In cases where no previous test results are available, most often a Bayesian method based on vague prior information is used. The assumption of a vague or uniform prior results in a conservative approach in cases where only a limited number of test results are available. However, in case of concrete, prior information on concrete strength can be found in literature or country-specific prior information can be determined. Therefore, the use of a combined vague-informative prior is of particular interest, more specifically in the form of scaled inverse-χ² distributions that can be used for updating the standard deviation of the strength distribution of concrete. The differences between the use of a vague and a combined vague-informative prior are investigated through Monte Carlo simulations. Because prior information is taken into account, the uncertainty regarding the standard deviation of the predictive strength distribution is significantly reduced.     

How attitude strength biases information processing and evaluation on the web

In this study we examined the influence of attitude strength on the processing and evaluation of sources and information on the Web. Seventy-nine German university students participated in an eye-tracking study in which they read information from eight pre-selected websites from different sources on the controversial topic of organic foods. Results showed that participants who felt strongly about the topic (i.e., those with strong prior attitude strength) scrutinized website logos of attitude-inconsistent websites shorter and judged the credibility of attitude-inconsistent websites lower. They also included more attitude-consistent information in an essay task than participants with weaker prior attitudes. Participants who felt less strongly about the topic (i.e., those with weaker prior attitudes) fixated longer on text from attitude-inconsistent websites than from attitude-consistent websites. By contrast, for participants with strong prior attitudes the time spent on the texts from attitude-consistent websites and attitude-inconsistent websites did not differ significantly. The results show that prior attitudes can bias evaluation and processing of information in different ways. Even though participants were not fully biased during initial information processing, they were so when evaluating the information and presenting it in an essay task.